ML Experiments

Central repository for playing around with new and exciting technologies

Setup

Certain experiments require setting environment variables, e.g., for authentication with services like Docker Hub, Paperspace, etc. See the .env.example file for all environment variables that may be required. The techonologies/datasets below list the required environment variables.

Technologies

List of useful resources for finding new technologies:

• Awesome Production ML: Github

List of technologies used for the experiments:

• NumPy
• pandas
• Matplotlib
• scikit-learn
• Lifelines
• pytest
• Hypothesis

Datasets

List of data sources:

• IEEE-CIS Fraud Detection: Kaggle competition
• Insurance cost prediction: Kaggle dataset
• Credit card customers: Kaggle dataset
• News API:
  • Open a free developer account to retrieve API key.
  • Add NEWS_API_KEY=[YOUR_KEY] to .env file.

Planned experiments

Unsupervised learning:

• Apply clustering techniques to credit card dataset
• Test t-SNE algorithm (see guide from Distill.pub)

Issues

• Docker image: Enable GPU support for TensorFlow
• Documenatation: Add more detailed setup instructions